-----
* chore: added py37 to setup.cfg
* test: add Python 3.7 to travis testing
* chore: add documentation and rename dijkstra to dijkstra3d
* chore: setting up for PyPI distribution
* test: fill out cpp test code with an example
* docs: update timestamp
* fix: euclidean\_distance\_field could be succeptible to non-boolean fields
* perf: faster dijkstra3d
* perf: faster euclidean distance field
* perf: 10% faster
* perf: bug in distance\_field3d was hurting performance a lot
* fix: rows/cols swapped in path\_from\_parents
* docs: updated module help
* perf: added parental\_field and path\_from\_parents
* docs: added anisotropy to euclidean\_distance\_field help
* feat: euclidean distance field
* feat: supported uints more fully for dijkstra
* doc: why does fortran order work? idk
* fix: simplified distance\_field
* fix: reconciled various transpositions
* test: added more complex test to dijkstra
* fix: algorithm was not properly checking visited
* fix: handle fortran order (by converting to C order..)
* docs: elaborated on where different contributions to memory usage come from
* docs: added benchmark
* feat: added support for uints and bools
* docs: added reference to Dijkstra paper
* docs: how to use distance\_field, performance info
* docs: added travis badge
* test: added 3d tests
* test: added automated tests
* fix: pyx handle Fortran ordered arrays in distance\_field
* fix: memory leak in distance field calculation
* fix: accomodate 2d and 3d inputs to dijkstra
* fix: row col depth assignment in distance\_field
* fix: corrected neighbor calculation and signed bit is actually needed
* feat: added prototype of distance\_field function
* docs: show how to use
* fix: made point cloud output sensible for 2D
* refactor: broke dijkstra function into components for neatness
* feat: running python code
* feat: Cython bindings for dijkstra3d
* feat: added dijkstra namespace, 2d and 3d functions
* refactor: make x,y,z translation more explicit
* perf: mostly eliminated compute\_neighborhood as a CPU suck
* perf: shaved another 10sec off the 512^3
* fix: removed pairing\_heap.hpp include
* perf: switched to std::priority\_queue, 20% speedup
* feat: moving to 26-hood
* feat: moved to 18-hood from 6-hood
* refactor: allocate heap instance on stack (not the full tree)
* refactor: allocate neighborhood on the stack
* fix: y coordinate computed incorrectly
* feat: now returns st path as a vector of array indicies
* perf: break out of the loop early, saves huge computation
* perf: a more dangerous, but highly performant improvement to running time
* perf: faster delmin through reserving vector size
* fix: dijkstra working but kinda slow, but very low memory
* fix: compilation warnings
* fix: improper deletion of heap
* fix: segmentation fault from accessing dist after free
* wip: moving towards a heap based dijkstra
* chore: add -std=c++11 to test compilation
* perf: why use two operations when you can have one
* fix: don't segfault when printing with no root
* perf: ensure compiler can optimize forest.size()
* fix: propogated "last" fix for odd sizes
* fix: pairing heap now sorts numbers
* docs: fixed formatting in a comment
* test: added heap sort test
* fix: construct PHNode correctly
* fix: syntax errors in pairing\_heap.hpp
* feat: untested pairing heap implementation
* docs: added paper references
* feat: a classic algorithm, O(V^2)
* Initial commit